Yi Yang

Elucidation and engineering of arabinofuranosyltransferase to enable total de novo biosynthesis of paris saponins in yeast

This paper reports the first complete de novo biosynthesis of four bioactive Paris saponins in yeast by engineering sugar donor specificity and integrating biosynthetic modules.

Multidimensional variation and population stratification across 8000 complete human centromeres

Over 8,000 gapless human centromeres reveal population-stratified satellite-array, inversion and haplotype variation, multiple candidate kinetochore sites in 16.8% of centromeres, and single-base substitution rates comparable to flanking pericentromeres.

Deciphering complete archaic introgression sequences in modern human genomes

The paper introduces ASMaid, a method to detect archaic introgression sequences in human genomes, revealing more complete and diverse archaic segments than previous methods.

PFM: perturbed flow matching for structure-based drug design

PFM generates 3D molecules that bind to protein targets with competitive structures and SOTA binding affinities, while significantly accelerating the generation process.

Field–Flow–Front mapping of breast cancer evolutionary dynamics in situ

This paper maps breast cancer evolution in situ using spatial transcriptomics and other data to show that tumor progression is driven by spatially coherent fronts rather than predefined clonal branches.

Incorporation of single-neuron projectome-based connectivity motifs enhances the cortex-specific performance of artificial neural networks

This paper develops a connectome-informed neural network algorithm that improves RNN and LLM performance by incorporating natural cortical connectivity motifs.

The Origin and Evolution of Protein Synthesis: A Co-Adaptation Flexible-Rigid Docking Model Based on First-Principles Reasoning

This paper proposes a model for the origin of protein synthesis through co-adaptation of flexible and rigid molecular structures in a prebiotic environment.

Learning residue-level context for modeling protein-protein interactions

This paper introduces ReCLIP, a transformer-based framework that models protein-protein interactions at the residue level, showing high accuracy in predicting mutation effects and peptide-MHC binding.

Rhythmic temporal structure organizes recurrent dynamics to support sequential working memory

This study shows that rhythmic input improves sequential working memory by organizing recurrent neural dynamics around a dominant frequency, enhancing temporal-order representations.

Architecture of the type II secretion system

The paper uses AlphaFold and cryo-EM to reveal the structure of the type II secretion system's pilus assembly platform and its organization across membranes.

EnzCast: Prediction of Patient-Specific Enzymatic Kinetics through Multi-Modal Deep Learning and Isoform-Resolved Bayesian Inference based on Single-Cell Transcriptomics

EnzCast predicts enzymatic kinetics from multi-modal data, and IsoKin infers patient-specific in vivo kinetics, showing improved uncertainty reduction in colorectal cancer.